This is the part that doesn't get talked about enough. Code review was never just about catching bugs it was how teams built shared understanding of the codebase. When someone skips reviewing their own AI-generated PR, they're not just shipping unreviewed code, they're opting out of knowing what's in their own system. The trust problem isn't really about the AI output quality, it's about whether the person submitting it can answer questions about it six months from now.